Validity and Reliability of the Inertial Measurement Unit for Barbell Velocity Assessments: A Systematic Review

The use of inertial measurement unit (IMU) has become popular in sports assessment. In the case of velocity-based training (VBT), there is a need to measure barbell velocity in each repetition. The use of IMUs may make the monitoring process easier; however, its validity and reliability should be established. Thus, this systematic review aimed to (1) identify and summarize studies that have examined the validity of wearable wireless IMUs for measuring barbell velocity and (2) identify and summarize studies that have examined the reliability of IMUs for measuring barbell velocity. A systematic review of Cochrane Library, EBSCO, PubMed, Scielo, Scopus, SPORTDiscus, and Web of Science databases was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. From the 161 studies initially identified, 22 were fully reviewed, and their outcome measures were extracted and analyzed. Among the eight different IMU models, seven can be considered valid and reliable for measuring barbell velocity. The great majority of IMUs used for measuring barbell velocity in linear trajectories are valid and reliable, and thus can be used by coaches for external load monitoring.


Introduction
Velocity-based training (VBT) is a resistance training method consisting of monitoring the velocity of movement displacement to support the regulation of load imposed on athletes [1][2][3][4][5]. Therefore, a proper measurement of bar displacement velocity is critical to implement an auto-regulation process of sports training [6][7][8][9][10]. Three reasons can be cited for using velocity as the main outcome [11][12][13]. First, there is a relationship between velocity and the amount of external mass lifted, by which a reduction in lifting velocity occurs as load increases until a terminal velocity is achieved at the maximal load [14]. Second, a nearly perfect linear relationship between velocity and intensity can be observed in many exercises and movements performed at different loads [15][16][17]. Third, reductions in voluntary exercise velocity are strictly related to neuromuscular fatigue induced by the exercise [18][19][20].
If an athlete is to benefit from VBT, certain instruments should be used to ensure that the velocity of movements is accurately and precisely measured [21,22]. For this purpose, different commercial devices can be used to quantify velocity [23]. Among the available options, solutions can be grouped as follows [24]: (i) isoinertial dynamometers consisting of a cable-extension linear velocity transducer attached to the barbell [25][26][27], (ii) optical motion sensing systems or optoelectronic systems [28][29][30][31], (iii) smartphone applications involving frame-by-frame manual inspections [29,32,33], and (iv) inertial measurement units (IMUs) [34]. Since these different technologies offer different possibilities, it can be considered that IMUs represent the most easy-to-use solution because no cable-extension is needed-the sensor simply needs to be attached to the barbell. Compared with video-based solutions, IMUs are also easier and quicker since no operations need to be made [35,36].
IMU solutions use fusion sensing to estimate velocity [37]. Thus, despite their practical benefits, some issues related to accuracy and precision should be considered. IMUs combine accelerometers (usually triaxial), a gyroscope (usually triaxial), and magnetic sensors to provide information about velocity, orientation, and gravitational force [35,38]. Despite the combination of sensors, there is always a margin of error related to the accuracy and precision of the estimations [39]. This margin of error should be understood so that better inferences can be made about human performance variability [40]. In fact, if validity or reliability is neglected, the results can be misunderstood, possibly affecting the judgments of coaches about their athletes [41][42][43][44][45].
On the basis of the importance of confirming the validity and reliability of IMU devices, different original studies have reported the results for different models in the sports sciences community [46][47][48][49]. Naturally, different experimental protocols have led to different results, and not all of the models are covered in the same conditions. Therefore, there is a need for a systematic review summarizing the validity and reliability levels of different IMU models during barbell movements. This will help us to understand whether coaches and athletes can use this technology to monitor resistance training that considers variations in human performance as opposed to in the devices [50].
While several systematic reviews have been published about the use of IMUs [50][51][52][53], no systematic review has summarized the validity and reliability levels of different IMU models for measuring barbell velocity. Considering the importance of the accuracy and precision level of determining barbell velocity in providing adequate prescriptions of resistance training, the aim of the present systematic review was twofold: (1) to identify and summarize studies that have examined the validity of wearable wireless IMU for measuring barbell velocity, and (2) to identify and summarize studies that have examined the reliability of IMUs for measuring barbell velocity.

Materials and Methods
The systematic review strategy was conducted according to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [54]. The protocol was registered with the International Platform of Registered Systematic Review and Meta-Analysis Protocols with the number 2020120135 and the DOI number 10.37766/in-plasy2020.12.0135.

Eligibility Criteria
The inclusion and exclusion criteria can be found in Table 1. The screening of the title, abstract, and reference list of each study to locate potentially relevant studies was independently performed by 2 of the authors (F.M.C. and M.R.G.). Additionally, they reviewed the full version of the included papers in detail to identify articles that met the selection criteria. An additional search within the reference lists of the included records was conducted to retrieve additional relevant studies. A discussion was made in the cases of discrepancies regarding the selection process with a third author (Z.A.). Possible errata for the included articles were considered.

Inclusion Criteria Exclusion Criteria
Test of a wearable wireless IMU. Instruments other than wearable wireless IMU.
Tests were conducted in barbell movements.
The tests were not conducted in barbell movements (e.g., human movements, other instruments).
Estimation of barbell velocity (m/s). Estimation of other outcomes than velocity (e.g., displacement).
In the case of validity, the IMU was compared with (i) an isoinertial dynamometer consisting in cable-extension linear position transducer, or (ii) optoelectronic system.
For validity, the IMU was compared with other instrument than isoinertial dynamometer or optoelectronic system (e.g., smartphone application; other IMU).
In the case of validity, one of the following measures were included: (i) typical error, (ii) mean absolute error, (iii) correlation coefficient, and (iv) standard error of the estimate.
For validity, outcomes presented are not typical error, mean absolute error, correlation coefficient, or standard error of estimate.
In the case of reliability, one of the following measures were included: (i) intraclass correlation test, (ii) coefficient of variation, (iii) standardized typical error and (iv) standard error of measurement.
For reliability, outcomes presented are not (i) intraclass correlation test, (ii) coefficient of variation, (iii) standardized typical error, and (iv) standard error of measurement.
Only original and full-text studies written in English.
Written in languages other than English. Article types other than original (e.g., reviews, letters to editors, trial registrations, proposals for protocols, editorials, book chapters, and conference abstracts).

Information Sources and Search
Electronic databases (Cochrane Library, EBSCO, PubMed, SPORTDiscus, and Web of Science) were searched for relevant publications prior to 1 January 2021. Keywords and synonyms were entered in various combinations in the title, abstract, or keywords: (sport* OR exercise* OR "physical activit*" OR movement*) AND ("inertial measurement unit" OR IMU OR acceleromet* OR "inertial sensor" OR wearable OR MEMS OR magnetometer) AND (Validity OR Accuracy OR Reliability OR Precision OR Varia* OR Repeatability OR Reproducibility OR Consistency OR noise) AND (barbell OR bar). Additionally, the reference lists of the studies retrieved were manually searched to identify potentially eligible studies not captured by the electronic searches. Finally, an external expert was contacted in order to verify the final list of references included in this scoping review in order to understand if there was any study that was not detected through our research. Possible errata were searched for each included study.

Data Extraction
A specific Excel spreadsheet was prepared for data extraction (Microsoft Corporation, Readmon, WA, USA) following the guidelines of Cochrane Consumers and Communication Review Groups [55]. The spreadsheet was used to identify the accomplishment of inclusion or exclusion criteria and to support the selection of the articles. The process was made by 2 of the authors (F.M.C. and M.R.G.) in an independent way. After that, they compared the results, and any disagreement regarding the eligibility was discussed until a decision was made in agreement.

Data Items
The following information was extracted from the included original articles: (i) validity measure (e.g., typical error, absolute mean error) and (ii) reliability measure (e.g., intraclass correlation coefficient (ICC) and/or typical error of measurement (TEM) (%) and/or coefficient of variation (CV) (%) and/or standard error of measurement (SEM)). Additionally, the following data items were extracted: (i) type of study design, number of participants (n), age-group (youth, adults, or both), sex (men, women, or both), training level (untrained, trained); (ii) characteristics of the wearable wireless IMU and comparator (isoinertial dynamometer consisting in cable-extension linear position transducer or optoelectronic system); (iii) characteristics of the experimental approach to the problem, procedures, and settings of each study.

Methodological Assessment
The STROBE assessment was applied by 2 of the authors (J.P.O. and M.R.G.) to assess the methodological bias of eligible articles following the adaptation of O'Reilly et al. [51]. Each of the included articles was scored for 10 items [51]. In cases of disagreement, it was discussed and solved by consensus decision. The assessment process was made in an independent way. After that, both authors compared the results, and any disagreement regarding the scores were discussed and made a decision in agreement. The study rating was qualitatively interpreted following O'Reilly et al. [51]-from 0 to 7 scores, the study was considered as risk of bias (low quality), whereas, if the study was rated from 7 to 10 points, it was considered as a low risk of bias (high quality).

Study Identification and Selection
The searching of databases identified a total of 159 titles (Cochrane Library = 11; EB-SCO = 59; PubMed = 31; SPORTDiscus = 31; Web of Science = 27). These studies, together with another two included from external sources, were then exported to reference manager software (EndNote X9, Clarivate Analytics, Philadelphia, PA, USA). Duplicates (71 references) were subsequently removed either automatically or manually. The remaining 90 articles were screened for their relevance on the basis of titles and abstracts, resulting in the removal of a further 55 studies. Following the screening procedure, 35 articles were selected for in-depth reading and analysis. After reading full texts, a further 13 studies were excluded due to not meeting the eligibility criteria ( Figure 1).

Methodological Quality
The overall methodological quality of the cross-sectional studies can be found in Table 2.

Methodological Quality
The overall methodological quality of the cross-sectional studies can be found in Table 2. Note: Provide in the abstract an informative and balanced summary of what was done and what was found (item 1); state-specific objectives, including any prespecified hypotheses (item 2). Give the eligibility criteria, and the sources and methods of selection of participants (item 3); for each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group (item 4); explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen and why (item 5); give characteristics of study participants (item 6); summarize key results with reference to study objectives (item 7); discuss limitations of the study, considering sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias (item 8); give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence (item 9); give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based (item 10).

Results of Individual Studies: Validity of IMU for Estimation of Barbell Velocity
Information of the validity levels obtained in the included studies can be found in Table 4. Some of the studies listed in Table 4

Results of Individual Studies: Reliability of IMU for Estimation of Barbell Velocity
Information of the reliability levels obtained in the included studies can be found in Table 5. Generally, the Barsensei model was the only model not being considered reliable for more than one article [27,46]. The remaining models presented evidence of reliability.

Discussion
This systematic review aimed to identify and summarize studies that have examined the validity of wearable wireless IMUs for measuring barbell velocity and identify and summarize studies that have examined the reliability of IMUs for measuring barbell velocity. The IMUs in this study were compared with gold standards and previously tested devices as reference systems (i.e., linear transducers [56,[58][59][60]67], a contact platform [69], the field computation method [24,61,[64][65][66]70], and a motion capture system).
IMUs were evaluated during movements generally geared toward strength training. The studies investigated in this review included the following movements: the back squat [1,23,24,34,49,58,64,66,69]; the bench press [34]; the hip thrust [58]; the bench throw [23]; the prone bench pull [60,67]; the countermovement jump [61]; the power snatch, clean, and jerk [48]; and the hexagonal barbell deadlift. Validity and reliability studies of IMUs during Olympic lifts are quite limited [61], and thus it is believed that IMUs should be tested for different parts of the Olympic lifts.

Validity of IMU for Estimation of Barbell Velocity
Twenty-one of the studies in this systematic review investigated validity (see Table 4). Specifically, IMU were compared with linear transducers in seven studies, a contact platform in six studies, the field computation method in one study, and a motion capture system in six studies (see Table 4 to a detail information). Among the included studies, nine tested the back squat; nine the bench press; one the hip thrust; one the bench throw; one the prone bench pull; two the countermovement jump; one the power snatch, clean, and jerk; and one the hexagonal barbell deadlift.
Validity studies also compared the different pieces of equipment with which IMUs were compared. The most detailed investigations are those tested with 3D camera measurement systems and force platforms, which are the gold standard. The variety of equipment used for validity comparisons can lead to differences. Differences in diversity between devices may be due to different sampling methods and the way raw data signals are processed in the software. Therefore, practitioners of IMUs should avoid using different devices interchangeably during the long-term monitoring of athletes.
The statistical methods and working designs of the equipment tested for validity differed. According to brands and methods for the BarSensei model used to test the validity of the findings of IMUs, the SEE values of validity were between 0.03 and 0.06 m•s −1 [46]. For the Gyko Sport model, the SEE and Pearson's r values were 0.18 m•s −1 and 0.79, respectively [49]. For the Beast Sensor model, the SEE values ranged between 0.07 and 0.05 m•s −1 , and Pearson's r values ranged between 0.76 and 0.98 [24,34]. For the Myotest sensor model, the SEE values ranged between 0.01 and 26.6 m•s −1 , Pearson's r values ranged between 0.38 and 0.92, and the R 2 values ranged between 0.59 and 0.97 [56,58,59,65,67,69]. For the PUSH Band sensor model, the SEE values ranged between 0.135 and 0.091 m•s −1 , Pearson's r values ranged between 0.97 and 0.90, and the R 2 values were around 0.85 [6,23,24,48,64,66]. For the Wimu RealTrack Systems sensor model, the SEE values were around 0.030 m•s −1 , Pearson's r values ranged between 0.009 and 0.60, and the R 2 values ranged between 0.95 and 0.77 [28,60,68]. For the PASCO sensor model, Pearson's r values ranged between 0.84 and 0.93 [61,70]. For the Barsensei model, the SEE values of validity were between 0.03 and 0.06 m•s −1 [46]. For the Gyko Sport model, the SEE values and Pearson's r were 0.18 and 0.79, respecitvely [49]. For the Beast Sensor model, the SEE values were between 0.07 and 0.05 m•s −1 and Pearson's r values were between 0.76 and 0.98 [24,34]. For the Myotest sensor model, the SEE values were between 0.01 and 26.6 m•s −1 , Pearson's r values were between 0.38 and 0.92, and R 2 values were between 0.59 and 0.97 [56,58,59,65,67,69]. For the PUSH Band sensor model, the SEE values were between 0.135 m and 0.091 m•s −1 , Pearson's r values were between 0.97 and 0.90, and R 2 value was 0.85 [6,23,24,48,64,66]. For the Wimu RealTrack Systems sensor model, the SEE value was 0.030 m•s −1 , Pearson's r values were between 0.009 and 0.60, and R 2 values were between 0.95 and 0.77 [28,60,68]. For the PASCO sensor model, Pearson's r values were between 0.84 and 0.93 [61,70].
Considering the scenarios in which instruments may not be recommended, we found that the Wimu and Myoset may not be appropriate for measuring countermovement jumps, while Barsensei is not recommended for measuring velocity in squat and back squat exercises. The Myotest, Push Bando, Wimu, and Pasco revealed validity for measuring the main weight-room exercises such as bench press, bench throw, squat (front and back), or deadlift. The experience or type of competitive level of the participants had no effect on the tests.
In light of the findings revealed in the systematic review, sports scientists and practitioners should question the validity of the IMUs they use during exercises. Even if they do not have appropriate conditions to validate UMIs, it is recommended that they use validated equipment as shown by the data discussed in this systematic review.
Reliability findings of IMUs, according to the brands and methods for the BarSensei model, the ICC values of validity were between 0.273 and 0.451, and the CV values ranged between 10% and 30% [27,46]. For the Gyko Sport model, the ICC value of reliability was 0.774 [49]. For the Beast Sensor model, the ICC values of reliability were between 0.36 and 0.99, and the CV values were around 35% [24,32]. For the Myotest model, the ICC values of reliability were between 0.35 and 0.97, the CV values were between 2.1% and 36.5%, and the SEM values of reliability were between 3% and 990 [56][57][58]65,67]. For the PUSH Band model, the ICC values of reliability were between 0.58 and 0.97, the CV values were between 4.2% and 13.7%, and the SEM values of the reliability were between 0.008 and 9.34 m•s −1 [23,24,48,64,66]. Depending on the number of studies, the PUSH BAND appears to be the IMU-based device that provides the most reliable data. For the Wimu RealTrack Systems model, the ICC values of reliability were between 0.81 and 0.97, the CV values were between 2.60% and 17%, and the SEM values of the reliability were between 0.007 and 0.11 m•s −1 [60,62,68]. For the PASCO model, the ICC values of reliability were between 0.95 and 0.99, and the SEM values of the reliability were between 0.55 and 1.77 m•s −1 [61].
In brief, the Myotest did not reveal enough levels of precision (reliability) for measuring countermovement jump, while Barsensei was not precise for measuring velocity in squat and back squat exercises. The Myotest, Push Bando, Wimu, and Pasco revealed precision for measuring the main weight-room exercises such as bench press, bench throw, squat (front and back), or deadlift. This is extremely important since instruments must be as must precise as possible in order to provide useful and sensitive information about readiness monitoring in sports, particularly for VBT.
The studies discussed in this systematic review investigated the reliability of the devices in certain movement patterns in field conditions. According to the authors, it is thought that studies on the long-term use of investigations of the reliability of IMUs should be designed to minimize the variables that might arise as a result of biological differences in longer use. However, it is also thought that malfunctions in the software data flow originating from the manufacturer may occur, and the disruptions in this data flow may affect the data reliability in IMUs. It is thought that research should be conducted to examine whether software and mobile phone applications that reflect instant data of IMU models transmit data reliably in real time. Among the studies discussed in this systematic review, none considered this situation. At the same time, to the best of the authors' knowledge, there is no study in the literature that investigates the effects of software data flows on reliability.

Study Limitations, Future Research, and Practical Implications
Most of the research in this systematic review examined the validity and reliability of IMUs during movements performed in a single plane. This systematic review only tested the validity and reliability of the Flores et al. [61] Olympic lifts and IMUs. Practitioners, sports scientists, and strength and conditioning coaches often use multi-directional Olympic lifts as opposed to the limited movement patterns used in research. Future studies should examine the validity and reliability of IMUs during the Olympic lifts that appear as a dark zone.
The studies discussed in this systematic review generally consist of short-term research designs. Since IMUs are used by people with biological differences, it should be considered that long-term biological changes may affect the validity and reliability of IMUs in the long term. To the best of the authors' knowledge, there are no studies examining the validity and reliability of long-term IMUs. For this reason, future studies should examine the validity and reliability of data to uncover insights about the long-term use of IMUs.
There may be some factors affecting the validity and reliability of IMUs beyond those that researchers have considered in their experimental designs. Some of these factors may be caused by the manufacturer. For example, it is believed that data transferred to software and mobile phone applications in real time result in errors in validity and reliability due to software malfunctions. In order for sports scientists, practitioners, and strength training coaches to use IMUs in a valid and reliable way, the effects of software factors on data quality should be investigated in future studies. Table 6 presents a summary of the validity and reliability of different IMUs that may help coaches choose a model.

Conclusions
This present systematic review summarized evidence about the validity and reliability of IMUs for measuring barbell velocity. A total of eight models were tested across the 22 included articles. The Barsensei was not valid and reliable in the studies reports. The Gyko sport, Beast Sensor, and PASCO were valid and reliable in all reports. The Myotest, PUSH band, and Wimu RealTrack were valid and reliable in the majority of the reports. The Rehagait was valid. Therefore, from the eight included models, seven can be used with some evidence of being accurate and precise. This evidence provides important information for coaches who need accurate information about barbell velocity to control the external load imposed on athletes and to be sensitive to human variations without any meaningful bias generated by measurement instruments. Funding: This work is funded by Fundação para a Ciência e Tecnologia/Ministério da Ciência, Tecnologia e Ensino Superior, through national funds and when applicable co-funded EU funds under the project UIDB/50008/2020. No other specific sources of funding were used to assist in the preparation of this article.